Silicon photonics has matured into a leading photonic integration platform [1,2]. Remarkable progress has been achieved in terms of device functionality, performance, and circuit integration for a wide range of applications, including telecommunications, datacom, sensing, and quantum photonics [3–7]. Optical interfaces play a key role in silicon photonics as they enable the efficient coupling of optical signals between integrated photonic devices and the off-chip environment [8]. In off-chip coupling scenarios, optical beam steering is an important technique due to its capability to control the direction of light propagation dynamically. Ideally, a beam steering system would be small and lightweight so that it can be mounted on a robot, an autonomous vehicle, a satellite, or integrated on a handheld device, such as a smartphone [9,10]. However, state-of-the-art optical beam steering systems typically comprise mechanical assemblies, moving parts, and bulk optic components [11,12].

On-chip optical phased arrays (OPAs) have gained significant interest as non-mechanical beam-steering devices capable of controlling optical wavefronts at high speeds [13–25]. A two-dimensional (2D) OPA typically incorporates an array of concatenated power splitters or a star coupler that distributes light to multiple waveguide channels. Each channel is equipped with an independent phase shifter, followed by an emitter element, i.e., a waveguide antenna or an edge coupler [17]. Surface grating couplers have been extensively used in silicon-based planar waveguides for fiber-chip coupling and wafer-scale testing [26–31].

In order to maximize the overlap with optical fiber mode and ensure high coupling efficiency, grating couplers are typically 10–15 μm long [32–35]. For 2D OPA applications, much shorter coupler lengths are desirable to make it possible to build a high-density array [36–41]. Moreover, to increase the beam steering range of a 2D OPA, the field of view (FOV) of the antenna must be wide enough to fully exploit the free beam steering range in OPAs with a small antenna pitch. In a 1D OPA (e.g., Ref. [11]), this issue can be readily addressed by narrowing the antenna (long grating coupler) transversally while the beam steering in the longitudinal direction is achieved using a wavelength sweep. On the other hand, improving antenna performance for the 2D OPA is fundamentally challenging. This issue arises because Fraunhofer diffraction dictates minimizing the antenna aperture to enlarge the FOV. However, this contradicts the fact that a considerable antenna length in the longitudinal direction is required to maintain sufficient coupling efficiency [42]. Our antenna overcomes this fundamental limitation by simultaneously yielding high efficiency and substantially broadened radiation patterns in the longitudinal direction. This is particularly significant for the development of large-steering-range OPAs.

Here, we report the first experimental demonstration of two fundamental theoretical concepts, namely near-field phase engineering and subwavelength transversally interleaved chirped structure. The proof-of-concept device is demonstrated for the first time, to the best of our knowledge, confirming the feasibility of the design strategy theoretically outlined in Ref. [43]. The antenna yields a record experimental performance, characterized by a far-field beam width of 44°×52°, an ultra-compact footprint of 3.4μm×1.78μm, and a coupling efficiency as high as −3.22dB. Furthermore, a novel topology of a 2D antenna array is demonstrated for the first time. This is the most compact 2D antenna array ever achieved for a comparable number of antennas (up to 20×8). An important advantage of this new topology is its scalability, which is achieved by leveraging an evanescently coupled architecture without compromising the compactness of the OPA. This novel topology also allows the freedom to apodize the optical waveform through the control of the coupling gap.

The far-field distribution of an antenna is intricately tied to its aperture size, governed by Fraunhofer’s transformation, which links the far field with near field characteristics. While diminishing the antenna length augments the field of view and steering range, reducing the aperture size inevitably compromises antenna efficiency. In order to circumvent this trade-off between the far-field beam width and antenna efficiency, we proposed a novel concept of near-field phase engineering [43]. Controlling the near field phase allows for substantially broadening the grating far field compared to conventional gratings. Our innovative antenna is implemented by transversally interleaving, at subwavelength pitch, two diffraction gratings with different periods, as shown in Fig. 1, resulting in a subwavelength grating (SWG) metamaterial in the transverse direction. SWGs, since their early demonstrations in silicon waveguides [44–49], have become a fundamental tool for controlling the electromagnetic field distribution in the integrated photonic devices [50,51]. In our antenna, SWG nanostructure allows for precise control of the near field phase, effectively overcoming the limit imposed by the Fraunhofer diffraction regime on the antenna’s far-field beam width. In this way, we succeeded in decoupling the antenna aperture size from the far-field beam width, which is required for achieving a specific scattering efficiency.

The antenna comprises a single-etch structure implemented in an SOI platform with a 300-nm-thick silicon waveguide core, a 1-μm buried oxide (BOX), and a 2-μm silicon dioxide cladding. The device is optimized for the transverse electric (TE) polarization. Each of the two interleaved diffraction gratings incorporates four periods along the longitudinal direction (Fig. 1, x-axis), which includes two fully etched sections with lengths L1 and L3 and two un-etched sections with lengths L2 and L4. These two longitudinal gratings are interleaved seventeen times in the transverse direction (y-axis) with a segment width of 105 nm, and therefore a subwavelength period of 210 nm, resulting in a total antenna width of 1.78 μm. The fundamental concept of the antenna and simulation results have been reported in our recent theory paper [43]. The underlying principle is to widen the antenna far field by carefully managing the phase of the near field.

By incorporating a particular near-field phase factor in the Fraunhofer transformation, we can expand the far-field beam width without reducing the antenna’s length. The near-field phase is controlled by transverse chirping, where two diffraction gratings of different periods are transversally interleaved at the subwavelength scale. As the different chirp periods equally contribute to the scattered field, this enables effective phase engineering in very short gratings (few periods), yielding a substantially broadened far-field radiation compared to the conventional antennas. This phenomenon can also be understood through the effect of longitudinal index engineering. The proper set of short grating periods forms a large equivalent period of up to a few micrometers, resulting in smaller longitudinal sections of different filling factors.

Synthesizing these new filling factors on the subwavelength scale will generate a new set of equivalent material indexes, which introduces efficient index chirping and eventually broadens the radiation pattern. For example, compared to longitudinally chirped grating coupler of three periods, where only three different diffraction angles can be chosen to build the phase contour, the index chirping in the transversally interleaved structure is able to bring richer phase components to widen the diffraction cone. Detailed specifications include etched and un-etched segments of the first grating that are L1=138nm, L2=66nm, L3=195nm, and L4=221nm, i.e., the period Λg1=620nm. For the second grating, the parameters are chosen as L1=114nm, L2=53nm, L3=194nm, and L4=392nm, yielding the period Λg2=753nm.

To further enhance the coupling efficiency, a Bragg reflector is added at the antenna’s end, recirculating the remaining power back into the antenna. The Bragg reflector comprises two periods of a silicon segment with a width of 153 nm, separated by 157 nm gaps, with a 305 nm separation from the grating. The use of a Bragg reflector also facilitates the widening of the far-field beam width in the longitudinal direction. This is achieved through the reversal of the diffraction angle, resulting in a notable increase of an additional 8° in the longitudinal beam width. The far-field beam becomes correspondingly wider, approximately 51°×47°, centered near −12° [43].

The devices were patterned using electron beam lithography, and the Si layer was fully etched by inductively coupled plasma reactive ion etching. The sample was coated with a 2.2-μm SiO2 layer through a plasma-enhanced chemical vapor deposition (PECVD) process and finalized, with added metallization for the phase shifters. The fabricated antenna is shown in Fig. 1.

Figure 2 shows the schematic of the experimental setup employed for characterizing the fabricated devices. We used the Agilent 81600B tunable laser with an external polarization controller to select in-plane TE polarization at the chip input. The light is coupled to the chip from a lensed fiber via an inverse taper edge coupler [52].

The optical setup loss was determined as 2 dB by sending the light from the optical fiber to the photodetector. The measured insertion loss of the edge coupler was −2.5dB. Figure 3 shows a comparative analysis of experimental and simulated results for the far field of a transversally interleaved chirped antenna. The angular distribution of the output light is recorded using a photodetector located approximately 10 cm above the chip. To restrict the angular range of incident light, a narrow-slit aperture was strategically placed in front of the photodetector, resulting in a scanning resolution of ∼1.4°. For the assessment of the FOV, the photodetector was mechanically rotated along the circumferential direction centered on the antenna, with a step size of ∼2°. The measured FWHM of the output power along the longitudinal (x) and transverse (y) directions is 44°×52° at the 1.55 μm wavelength, which agrees well with the theoretical value (49°×47°) shown in Fig. 3. The simulated results also confirm that this nominal configuration yields a substantially broadened far-field radiation pattern compared to a conventional grating (without phase engineering) of the same size (36°×48°).

Figure 4 compares the 3D-FDTD simulations and measurements of the upward diffraction efficiency of the antenna at λ=1.55μm. The measured efficiency follows a similar trend to the simulated results. However, we find that the measured wavelength of the peak diffraction efficiency is blue shifted by about 25 nm from the theoretical estimate. This can be attributed to fabrication bias, resulting in deviations from the nominal grating duty cycle. The peak efficiency at λ=1.55μm obtained from the simulation is −2.66dB, and the highest measured efficiency is −3.22dB at λ=1.525μm. Regarding antenna fabrication tolerance, we assessed changes in diffraction efficiency for variations in segment lengths. Our antenna demonstrates robust performance with less than 6% reduction in diffraction efficiency for a fabrication bias of 10 nm, achievable using state-of-the-art silicon photonics manufacturing. This robustness is attributed to the utilization of a subwavelength structure based on the effective index averaging effect.

A small proof-of-concept two-dimensional (2D) phased array was designed to showcase the antenna performance. Figure 5 shows the SEM image of the fabricated 2×4 array with the transversally interleaved chirped antenna elements, independently controlled by individual phase shifters. The antenna center-to-center spacing is 4 μm longitudinally and 2.2 μm transversally. The silicon thermo-optic (TO) effect, with a TO coefficient of ∼1.8×10−4K−1 at 1.55 μm [53], is used to generate the phase shift required for beam steering. The TO phase shifters are implemented using 700-μm-long tungsten-titanium (TiW) heaters deposited on the upper SiO2 cladding. The heaters are laterally spaced by 16 μm to ensure minimum thermal crosstalk and are connected to a thick bilayer of TiW/aluminum bond pads and subsequently linked to a printed circuit board (PCB) through precise wire bonding. The power supply for the phase shifters is facilitated by a current driver with digital-to-analog converters (DACs) and a field-programmable gate array (FPGA). Each DAC channel is capable of delivering electrical power of up to 200 mW.

For the OPA measurements, the light coupled to the chip is distributed to the waveguide arrays using Y-junction splitters. The schematic of the experimental setup is shown in Fig. 6. A three-lens measurement configuration flexibly facilitates the near-field and far-field image acquisition. An aspheric lens M1 (NA=0.66) and two achromatic lenses, M2 and M3, with the respective focal lengths of 15 mm, 75 mm, and 100 mm, are used alongside with a near-infrared InGaAs camera to generate and acquire the field images. Ray traces for the far-field and the near-field measurements are conceptually illustrated in Fig. 6. By controlling the working distance from the sample to lens M1 along with the camera position, either the far-field or the near-field pattern is selectively chosen to be imaged on the camera. For near-field observation, the working distance extends to 2f1 (twice the focal length of lens M1), ensuring that the image plane is positioned within the focal length (f2) of the second lens M2. The near-field image is then captured by the camera through lens M3. For the far-field observation, the working distance is f1, such that the far field formed by lens M1 is located at the focal plane of lens M2. The far field is then re-imaged and captured by the camera located at the focal plane of lens M3.

Figure 7(a) shows the measured far-field image of a 2×4 phased array at 1.55 μm wavelength, prior to compensation of the random phase offsets induced by fabrication imperfections. The far field is then calibrated by maximizing the power toward the zenithal direction filtered in the Fourier domain with a pin hole placed in front of the camera. We implement the calibration process using the genetic algorithm performed in the FPGA and a phase-power lookup table preliminarily established from a Mach–Zehnder interferometer with the optical phase shifter in one of the arms. After the calibration, we obtain the phase corrected far-field pattern shown in Fig. 7(b). The OPA far-field profile is also simulated [in Fig. 7(c)] using multi-wave interference calculation and the antenna radiation profile simulated with 3D-FDTD, showing a good agreement between the simulation and the experimental array far field results.

Figure 8 shows a comparison between the measured and simulated far field obtained at a maximum steering angle in horizontal and transversal directions for a 4μm×2.2μm pitch antenna array. By using Eq. (1) for Δϕ=π relating the uniform phase difference (Δϕ) between the diffracted beam from the antennas to the steering range (θ) [54], a simulated value of 22º × 41.25º, shown in Figs. 8(b) and 8(d), is calculated. This is consistent with the experimentally obtained maximum steering range of 19.3º × 39.7º. The small difference comes from the measurement inaccuracy of the distance between the optical components (chip, lens, and camera), introducing deviations in the pixel-to-angle calibration, θ=2arcsin(Δϕλ2πd),(1)where d is the spacing between the antennas, λ is the wavelength, and Δϕ is the phase difference.

The measured far-field profile of the steered beam is shown in Figs. 8(a) and 8(c), where a phase difference of Δϕ=π is applied between the antennas along the x- and y-directions, respectively. As can be seen, the beam is shifted by the full grating lobe distance from the main peak. The field profile steered along the x-axis, as shown in Fig. 8(a), and the original distribution in Fig. 7(b) (Δϕ=0) indicate that the antenna has a broad FOV along the x-direction, and the factor limiting the array’s steering range is the pitch in the x-direction. On the other hand, as shown in Fig. 8(c), the antenna’s FOV in the y-direction is close to the maximum accessible steering range determined by the array factor. The sidelobes and background noise in the center of the field likely arise from the incomplete destructive interference due to the opposite antenna orientations in a back-to-back configuration, as shown in Fig. 5.

The proof-of-concept design depicted in Fig. 5, featuring a 2×4 phased array, serves as a demonstration of the successful integration of our antenna into a compact array configuration. To address the challenge of scaling up the integration of the antenna while preserving the compactness of the phased array and avoiding the introduction of the grating lobes, we propose a second design showcased in Fig. 9(a). This array design comprises 8 rows of antennas, with each row accommodating 20 antennas fed by a common bus waveguide. The evanescent coupling of light from the bus waveguide to the antenna is a crucial aspect of this design. The antenna pitch is 4.2 μm and 2.7 μm, for x- and y-directions, respectively. The SEM image of the detailed structure is shown in the inset of Fig. 9(a). By judiciously tilting the antennas, the fundamental TE mode is coupled from the bus waveguide to the slab waveguide feeding each antenna. To ensure light is uniformly distributed to the antennas, the coupling gap between the bus waveguide and the antennas is gradually narrowed down along the propagation direction (x), as optimized using 3D FDTD simulations. At the end of the feeding waveguide, approximately 20% of the input light remains and is used for inspection and analysis. Each bus waveguide is routed out and equipped with a TO phase shifter for beam steering.

Figure 9(b) illustrates the measured uncalibrated OPA far field. Subsequently, we apply the phase calibration method to eliminate the fabrication-induced phase offset and obtain a calibrated far-field pattern, as shown in Fig. 9(c). Furthermore, the beam is steered by half of the grating lobe distance upon introducing a π phase shift between antennas, as illustrated in Fig. 9(d). The maximum steering range along the y-direction is 31.4º, limited by the 2.7 μm antenna pitch and antenna FOV. The fluctuations at the center of the image in Fig. 9(d) likely arise from fabrication imperfections in the Y-branch power splitters. The phase alignment along the x-direction is implemented by judiciously selecting the x-pitch and the mode effective index, specifically by choosing the waveguide width such that the x-pitch yields zero relative phase (modulo 2π) between the antennas on the same row. The presence of a noisy background likely results from scattering and the non-uniformity of the waveguides and splitters due to fabrication imperfections. Notably, scattering arises from the fabrication-induced tolerances for the small coupling gap, which significantly contributes to both the background noise and the splitting of the beam in Fig. 9(c). The choice of such a small coupling gap aims at effectively coupling the power from the bus waveguide to the array consisting of 20 antennas. This loss can be reduced by scaling up the array size, incorporating more antennas, hence a larger coupling gap. Currently, the beam steering on this design is only shown in the y-direction for an initial demonstration of the antenna integration. The beam steering in the x-direction can be readily implemented using the conventional method, including thermo-optic, plasma dispersion, and electro-optic effects.

In this paper, we experimentally demonstrate for the first time, and to the best of our ability, an ultra-compact waveguide antenna based on a transverse chirping strategy to substantially broaden the antenna far-field pattern. Our measurement results yield a broad far-field beam width of 44°×52° at 1.55 μm wavelength for an antenna aperture of 3.4μm×1.78μm, with a peak diffraction efficiency of −3.2dB at 1.52 μm wavelength. This outstanding performance is obtained with a single-etch grating nanostructure implemented in a 300-nm SOI platform. We also show the practical functionality of this novel antenna within simple phased array systems, providing a proof-of-concept demonstration. The experimental steering range of a 2×4 array is 19.3º × 39.7º at 1.55 μm wavelength, closely matching the simulation results. Furthermore, we demonstrate the scalability of the array configuration to 8×20 elements with a transverse steering range of 31.4º, leveraging new evanescently coupled antenna array topology. We believe that these results will pave the way for an innovative research direction in the development of highly-efficient nanophotonic antennas for on-chip optical phased arrays.